基于家族优生学的进化算法

A NEW EVOLUTIONARY ALGORITHM BASED ON FAMILY EUGENICS

模拟进化有几种典型方法，分别强调自然进化过程的不同侧面．人们已意识到恰当地结合这些方法将推进该领域的研究．本文将现有进化方法的差别归结为“遗传链”与“进化链”之间的差别，提出一种新的进化模型，称之为“基于家族优生学的进化ＦＥＢＥ（ｆａｍｉｌｙｅｕｇｅｎｉｃｓｂａｓｅｄｅｖｏｌｕｔｉｏｎ）”．新的模型用家族优生学的思想将遗传链与进化链溶为一体，将正交设计技术引入家庭的子代培植以加强个体的行为改进．本文将ＦＥＢＥ模型用于求解困难的Ｇｏｌｄｂｅｒｇ欺骗问题。

Several popular approaches of simulated evolution have been developed separately. These approaches emphasize different facets of the natural evolutionary processes, respectively. One has recognized that the simulated evolution will benefit from the adequate combination between the approaches. This paper characterizes the primary difference among existing approaches as the difference between genetic link and behavioral link. A new model of simulated evolution, called FEBE(family eugenics based evolution), is proposed, which combines the genetic link with the behavioral link in light of the idea of family eugenics. In the FEBE model the orthogonal design technique is introduced into offspring's breeding inside a family so as to enhance the behavioral improvement of individuals. The FEBE model is applied to solve Goldberg's deceptive problem that is challenging to most evolutionary algorithms. The exciting experimental results are achieved.